Multi-sourcing a single raw material from multiple predecessors

[rusk-ey]

When a node has 2 or more predecessors with the same upstream product, running the simulation results in an error.

For example, in an network defined by:

network = network_from_edges(
edges=[(2, 1), (3, 1)],
node_order_in_lists=[1, 2, 3],
local_holding_cost=[2,4,2],
stockout_cost=[50, 25, 25],
demand_type='UD', # discrete uniform distribution, for easier debugging
lo=1,
hi=5,
shipment_lead_time=[1, 1, 0]
)

nodes = {n.index: n for n in network.nodes}

products = {10: SupplyChainProduct(index=10), 20: SupplyChainProduct(index=20)}
products[10].set_bill_of_materials(rm_index=20, num_needed=3)

nodes[1].add_product(products[10])
nodes[2].add_product(products[20])
nodes[3].add_product(products[20])

products[10].inventory_policy = Policy(type='BS', base_stock_level=10, node=nodes[1], product=products[10])
products[20].inventory_policy = Policy(type='BS', base_stock_level=12, node=nodes[2], product=products[20])
products[20].inventory_policy = Policy(type='BS', base_stock_level=22, node=nodes[3], product=products[20])

I get the error... :

Traceback (most recent call last):
File ..., line 157, in
total_cost = simulation(network, 100, rand_seed=17)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ..., line 115, in simulation
step(network=network, consistency_checks=consistency_checks)
File ..., line 253, in step
_generate_downstream_orders(n.index, network, t, visited, order_quantity_override=order_quantity_override)
File ..., line 431, in _generate_downstream_orders
rm_OQ = order_quantity_dict[p_index][rm_index]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^
KeyError: -5

[LarrySnyder]

@bruskey47 Are you using the most recent commit? I get a different error message than what you're reporting!

[rusk-ey]

Here is the correct code for the error I reported - sorry for the confusion! The change is in the line:
products[10].set_bill_of_materials(rm_index=20, num_needed=3) ---> products[10].set_bill_of_materials(raw_material=20, num_needed=3)

network = network_from_edges(
edges=[(2, 1), (3, 1)],
node_order_in_lists=[1, 2, 3],
local_holding_cost=[2,4,2],
stockout_cost=[50, 25, 25],
demand_type='UD', # discrete uniform distribution, for easier debugging
lo=1,
hi=5,
shipment_lead_time=[1, 1, 0]
)

nodes = {n.index: n for n in network.nodes}

products = {10: SupplyChainProduct(index=10), 20: SupplyChainProduct(index=20)}
products[10].set_bill_of_materials(raw_material=20, num_needed=3) ##changed here

nodes[1].add_product(products[10])
nodes[2].add_product(products[20])
nodes[3].add_product(products[20])

products[10].inventory_policy = Policy(type='BS', base_stock_level=10, node=nodes[1], product=products[10])
products[20].inventory_policy = Policy(type='BS', base_stock_level=12, node=nodes[2], product=products[20])
products[20].inventory_policy = Policy(type='BS', base_stock_level=22, node=nodes[3], product=products[20])
total_cost = simulation(network, 100, rand_seed=17)

[LarrySnyder]

@bruskey47 The problem is that you are setting products[20].inventory_policy twice, and the second assignment overwrites the first assignment. So when it finds the order quantity at node 2 it gets confused because the inventory policy at node 2 thinks it lives at node 3 because of the node=nodes[3] argument.

To fix it, replace with:

nodes[2].inventory_policy = {20: Policy(type='BS', base_stock_level=12, node=nodes[2], product=products[20])}
nodes[3].inventory_policy = {20: Policy(type='BS', base_stock_level=22, node=nodes[3], product=products[20])}

(Setting inventory policies is definitely annoying and non-intuitive. This also relates to #162.)